Fifty years of R.

Bradshaw
C. Schroeder
IBM innovation
with information
storage on
magnetic tape
On May 21, 1952, the International Business Machines
Corporation announced the IBM Model 726 Tape Unit with
the IBM Model 701 Defense Calculator, marking the transition
from punched-card storage to digital storage on flexible
magnetic tape. That bold introduction was the beginning
of what is now a 50-year history of invention that has seen
remarkable advances in the storage of information on flexible
magnetic ribbons ten times thinner than a human hair and
capable of storing more than 100 000 times more data in the
same volume as the first reel of tape introduced in 1952. This
historical perspective is dedicated to the people who made
that first tape drive possible and to those who continue that
tradition in the Removable Media Storage Solutions (RMSS)
team of the International Business Machines Corporation Tape
Development Laboratory, headquartered in Tucson, Arizona.

How it all began and positioned figures within IBM were convinced that it
In June 1949, reels of tape, tape cartridges— even was foolish to risk the company’s future on such a radical
programmable computers— had yet to be invented. and unproven technology.
Recorded music was still on 78-rpm platters, and wire Wayne Winger, a member of IBM’s first tape-drive
recorders were the mainstay of the radio industry. development team and a Tucson retiree, recalls, “Once a
Magnetic tape had only recently emerged from the white-haired IBM veteran in Poughkeepsie pulled a few of
laboratory to be used for a few broadcasts. Meanwhile, at us aside and told us, ‘You young fellows remember, IBM
IBM, it took a large room full of vacuum tubes and miles was built on punched cards, and our foundation will
of wire to accomplish what a handheld calculator could do always be punched cards’ ” [1].
faster by the end of the 1980s. Information storage meant Thomas J. Watson Jr., then IBM executive vice
books, filing cabinets or, to those at the leading edge of president, and a number of the company’s top engineers
data processing technology, punched paper cards. A and scientists decided that the move to magnetic tape was
handful of IBM engineers and scientists were about to a gamble that had to be taken. Today’s multi-billion-dollar
launch a revolution. data-storage-processing industry is testimony to their
In 1949, it was by no means clear that it would be wise wisdom and their courage.
to gamble millions of dollars and the time of some of the In a documentary film, The 701 Days, produced by IBM
brightest engineers in America (to say nothing of the in 1973, the opposition to magnetics is recalled by people
future of the 35-year-old IBM Corporation) on an infant who participated in the climactic meeting of the IBM
technology based on gluing bits of rusted iron on strips of decision makers. In the film, IBM Fellow Nathaniel
plastic. The suggestion that such a vaguely understood Rochester recalled that Watson “went all around the room
thing, called magnetics, could supplant the familiar and asking people if this was the right thing to do or not, and
highly profitable punched card met strong and vocal some people said ‘yes’ and some people said other things.
opposition within the company. Some highly respected And then he told all those people who said other things

娀Copyright 2003 by International Business Machines Corporation. Copying in printed form for private use is permitted without payment of royalty provided that (1) each
reproduction is done without alteration and (2) the Journal reference and IBM copyright notice are included on the first page. The title and abstract, but no other portions, of this
paper may be copied or distributed royalty free without further permission by computer-based and other information-service systems. Permission to republish any other portion of
this paper must be obtained from the Editor. 373
0018-8646/03/$5.00 © 2003 IBM

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER
 meetings and discussions involving principally Gordon
Roberts and Steve Dunwell of the Future Demands
department, W. W. McDowell, vice president of
engineering, R. L. Palmer, manager of the Poughkeepsie
lab, and yours truly” [1].
A number of people remember that Tom Watson Jr.
immediately grasped the promise of magnetic recording as
a result of his experience with electronics in the Army Air
Corps during World War II. Why did Watson and a few
others decide that IBM had to commit itself to magnetic
tape? Phelps remembered it this way in his letter: “As I
recall, the major push first came from our study of the
Social Security System and their very pressing need to
solve their record storage problem. Even in 1949 they
had acres of file cabinets of IBM cards containing Social
Security records on every working American employee.
“It was obvious that we had to find a more compact
means of storing permanent records. And, there was not
much choice. We had tried more dense paper cards
(smaller holes, binary codes, etc.), but the projected
improvement was only a few times better. Punched paper
tape offered no great advantage either. Magnetic tape was
just coming into its own in audio systems and it offered
the most promise, so that’s the way we went.”
Development team member Wayne Winger recalled this
perspective on the limitations of punched cards: “Prior
to our first tape drive, information was put on punched
Figure 1 cards. There are 80 characters per card, and a good
Early laboratory prototype (a quarter-inch tape machine) of the IBM [read/write] speed was 100 cards per minute, which means
tape drive in the Kenyon House, Poughkeepsie, New York. 133 characters per second. The 726 (IBM’s first marketed
tape drive) operated at 7500 characters per second, which
is 56 times faster. And the tape takes up far less space.”
that they should work on other problems.” In the film, In 1949, the whole subject of magnetic recording was
Watson himself says, “Some of us decided we just had to virtually unexplored territory. Max Femmer, another tape
force ourselves into magnetic tape. And I wanted to get pioneer, recalled that “stuff you can learn in a primer
everyone’s views out on the table as to who was going to from Sam’s Bookstore today, we didn’t know.” 1 As in any
fight against it. And of course, when you have a meeting creative effort, improvisation played a large part in the
like that and you state that your position is going to be success of the tape drive development program. James
for it, it is helpful. . . .” Here Watson chuckles at the Weidenhammer, a key figure who invented the vacuum
recollection [1]. column, remembered that “we had no clear idea of which
Opponents of the magnetic tape program had well- approach to take. Very rapid start and stop times were
developed arguments against it. Fortunately, those obviously desirable to minimize the wasted tape and
arguments proved wrong. For example, Byron Phelps, a read/write delays. With tape speeds contemplated in the
principal figure in the development of magnetic tape and order of 100 to 200 inches per second, it was obviously
the holder of a number of key patents, wrote in a personal impractical to accelerate bulky tape storage reels rapidly
letter, “I can remember someone preparing a study on enough, so it was evident that storage loops or buffers
possible tape density and producing a graph showing that would be necessary in the tape path for gradual
400 bits per inch was just about the ultimate recording acceleration of the reels” [1].
density we could hope for!” (Today, the IBM Ultrium † Once the idea of the vacuum column was hit upon and
Linear Tape-Open † (LTO † ) Tape Cartridge records at the vacuum switch was conceived, “We were in a hurry to
over 124 000 bits per inch (bpi), and densities of over try out the idea,” wrote Weidenhammer, “and needed
250 000 bpi seem feasible in the very near future.)
“No one person came up with the idea of magnetic 1 Personal interview with Max Femmer in 1979 in Tucson, Arizona, with Carl
Schroeder of IBM in preparation for the IBM General Products Division Tucson
374 tape,” according to Phelps. “It grew out of a number of site dedication.

R. BRADSHAW AND C. SCHROEDER IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003
some very thin, flexible material in order to fabricate a
sensitive pressure sensing diaphragm. Nothing suitable
being on hand, the quickest solution that occurred to me
was to send one of the young engineers, Jack Seely, to the
nearest drug store for a pair of baby pants. They worked.”
The vacuum pump provided another challenge. During
development, engineers used the motor from an old
General Electric** vacuum cleaner, and released the tape
drive to production with that GE** motor. Dick Whalen,
who was manager of procurement for the program, wrote,
“It turned out that this was an obsolete vacuum cleaner.
A GE sales representative searched GE warehouses all
over the country to get us under way.” One of the early
prototypes that was used to work out a practical tape
processing machine is shown in Figure 1.
Wayne Winger recalled the confusion that resulted Figure 2
in the early days because magnetic tape was virtually
The IBM Model 726 Tape Unit, announced and shipped in 1952.
unknown, even within IBM: “We got a shipment of tape
from our supplier, and it was delivered to the receiving
dock. The man in charge came over and said, ‘We just got
a shipment of tape from 3M, but we’re going to have to
send it all back.’ Why? ‘It doesn’t have any glue on it!’ ” designed and built it—an indication of the pivotal role
IBM pioneered not only tape drives, but the half-inch played by magnetic tape in launching the data processing
tape itself. Vic Witt, who retired as an IBM Fellow in industry [2].
1980, played a large part in that program. “I started with The Wall Street Journal article on the day the 701 was
the program in 1951,” recalled Witt. “There were no tape announced said it was “designed to shatter the time
experts at that time. They handed me a piece of tape and barrier confronting technicians working on vital atomic
said, ‘You have to know everything there is to know about and airplane projects.” It went on to say that the first
this stuff.’ It was quarter-inch, and we were going to use computer would be used to “calculate atomic radiation
half-inch, but this was all we had then. effects and to compute the many statistical things
“We settled on 3M as a source for tape, and worked scientists need to know about planes, guided missiles,
with them to find a way to produce tape to our and rocket engines.”
specifications. We went to them and said, ‘We need tape The General Motors** Research Laboratories used a
so flawless that, if it were a highway, it would stretch from 701 in the mid-1950s, for example, to solve engineering
Poughkeepsie to Manhattan (about 60 miles) with no problems. One problem in stress analysis was evaluated
defects [larger than] the size of marbles.’ We built a test for more than 100 000 variables. A GM** publication
center in Minneapolis, and 100 percent of our tape was noted at the time that “working 40 hours a week, 52
inspected by an operator with a microscope and a knife. weeks a year with a desk calculator, it would have taken
Defects would be cut out and sent back, and we’d say, ‘Fix
the engineer 12 years to solve this problem. After allowing
that.’ Testing the tape cost as much as manufacturing it,
the mathematicians in Data Processing about a month to
maybe more.
set up the problem, the 701 provided the required answers
“IBM designed and built the world’s most advanced
in 1.5 hours.” As a GM employee noted, “Jobs once
tape coater in Poughkeepsie, and the first clean room used
thought impossible are now attacked without hesitation.”
in manufacturing. No one had ever heard of keeping the
The March 27, 1953, New York Times reported that one
air in a manufacturing plant that pure, but we had to have
it if the tape was going to meet our specifications. We chemical company planned to rent time on the 701 at
showed the world that it could be done” [1]. IBM headquarters to solve a cost-accounting problem. It
IBM’s first magnetic tape unit, the IBM 726 (Figure 2), would take about 100 hours of machine time to solve the
was announced in 1952—the year Thomas J. Watson Jr. problem while, according to an IBM spokesman at the
became IBM president and the employee population time, it would take an accountant with a desk calculator
passed 40 000. It was announced with the IBM 701 2500 years. The data processing industry’s first half-inch
Defense Calculator, the company’s first commercially magnetic tape drives, IBM 726 Tape Units, were shipped
marketed electronic computer. A development model was with the IBM 701 Defense Calculator from December 20,
called the Tape Processing Machine by the engineers who 1952 until February 28, 1955 [1, 2]. 375

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER
 surface, they are attracted to the recorded magnetic
regions on the tape and oriented in a manner that reflects
incident light differently depending upon their magnetic
alignment. This produces a visible pattern for the
alternating erased and aligned regions on the tape.
Figure 4 presents the image of a similarly developed
pattern for the servo and data bits on an IBM Ultrium
LTO 100-GB cartridge compared with a human hair, as
done for a similarly developed image for IBM’s first tape
drive, the Model 726.

Tape grows up
IBM’s magnetic tape development has been one of its
more-traveled missions. It started in Poughkeepsie, New
York, in the late 1940s and, for the first years, was housed
Figure 3 in the old Kenyon Mansion, which became the IBM
Homestead in 1955 and later served as the home of the
Developed bits on first IBM Model 726 100-bpi tape compared with
a human hair. Corporate Headquarters Management School—Northeast.
The tape mission then moved from the Kenyon Mansion
to the new Poughkeepsie Laboratory in the early 1950s.
The Poughkeepsie Laboratory developed refinements of
the early 727 Tape Unit, resulting in the 729 family of
tape drives, Models 1– 6. These drives offered increased
LTO 100 GB
384 tracks/0.5 inch density of up to 800 bpi with the introduction of two-gap
read/write heads, the first read-while-write capability in an
IBM tape drive, and the beginning of what became normal
tape operation for the next 25 years—write wide, read
narrow. These drives also pioneered the ability of an IBM
device to read and write multiple densities, making it
possible to read tapes written on older machines. As the
role of tape in digital information processing and storage
180 X
grew, IBM’s laboratory grew as well. In the early 1960s,
significant advances were developed in the Poughkeepsie
Laboratory that culminated in the 2401 family of tape
Figure 4 drives, first introduced in 1964 (Figure 5). Nine-track
non-return to zero inverted (NRZI) recording—a method
Developed data pattern for IBM 3580 Ultrium LTO 100-GB tape
of encoding binary data that offers data and clock
compared with a human hair.
synchronization advantages—was introduced with the 2401
Models 1–3, which added the capability of handling an
increased character set within the encoded data. Cyclic
Even at the first working density of 100 bpi of a half- redundancy checking (CRC) was introduced to provide a
inch tape, a 10.5-inch-diameter reel of tape could hold the much improved method of error checking and correction.
equivalent of more than 35 000 punched cards. Today, the The ability to read backward rather than force a rewind
IBM Ultrium LTO Tape Cartridge is 1000 times faster and restart allowed improved throughput and efficiency.
and stores more information on one inch of tape than the The final member of the 2401 family, the Models 4 – 6
IBM 726 Tape Unit was capable of storing on over 500 introduced in 1965, provided a top density of 1600 bpi and
feet of tape . . . without data compression. It is interesting saw the introduction of electronic skew buffers to reduce
to compare this early format with tape today. Figure 3 off-track errors and write compression to control peak
shows the bits recorded on a tape written on this early shift. The 2401 family of tape drives were to be the final
100-bpi tape drive compared with a human hair. The contribution to tape development from the Poughkeepsie
magnetized region of the tape is developed using a Laboratory, and the tape mission moved again—this time
dispersion of small iron particles (ferrofluid). When the to the new laboratory being constructed in Boulder,
376 suspended magnetic particles are rinsed across the tape Colorado. Then, in 1973, the tape mission moved again,

R. BRADSHAW AND C. SCHROEDER IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003
from the Boulder, Colorado, Laboratory to San Jose,
California, but it returned to Boulder in 1977.
While the tape mission was moving from Colorado to
California and back, the IBM tape team developed and
introduced some of the finest enhancements to reel-to-
reel tape devices, whose legacy had begun some 25 years
earlier, first with the 2420 Model 7 in 1968 with 1600 bpi
and a tape speed of 200 ips. This device exhibited the
fastest start–stop time of its day, going from a dead stop
to 200 ips in less than two milliseconds. The introduction
of high-torque, low-inertia (HTLI) motors coupled directly
to the drive capstan in the Model 2420 family of tape
drives enabled successive performance improvements in
the models that followed [3].

The media
Initially, the tape was produced by the 3M** Company
and several other manufacturers of audio recording tape.
The far more demanding performance and durability
required of tapes for computer data storage soon forced
IBM to put considerable effort into developing tape
specifically for IBM tape storage devices. The first
significant change was to replace the acetate substrates
used in earlier tapes with polyethylene terephthalate
Figure 5
(PET) to improve the tape’s resistance to breaking during
the increasingly faster start–stop cycles of each successive The IBM Model 2401 Models 1–3, 800-bpi nine-track (1964); this
was followed by Models 4–6, 1600 bpi, introduced in 1965.
model of tape drive [3].
In 1959, IBM introduced Durexcel, its first custom-
designed tape. Durexcel provided greatly improved
durability for the 726 and 727 Model tape drives. The The IBM 3850 Mass Storage Subsystem
IBM 729 Model III, with the introduction of pairs of The IBM 3850 Mass Storage Subsystem (MSS) was a
three-position assemblies which moved pinch rollers total departure from the 20-plus years of reel-to-reel
that allowed the tape to be engaged by the appropriate tape transports. It was a truly revolutionary product
capstans, called prolays, allowed improved start–stop that introduced information storage concepts and novel
performance and a significant reduction in mechanical technologies that appear to have been reinvented in the
complexity. These prolays, while providing more aggressive years that followed. The MSS was designed and built
tape performance, required the development of an at the IBM Boulder Laboratory in the early 1970s and
improved tape, introduced in early 1961 and called the introduced in 1974. It was the first attempt to virtualize
Heavy Duty Tape. Almost all of the early tapes were storage by embedding software instructions within the
produced by suppliers to IBM, but the need to optimize device to control its functions; these instructions are now
tapes specifically to support the requirements of computer commonly known as microcode. The 3850 MSS used a
data storage devices necessitated that IBM design and helical-scan format on tape to mimic data as it would
manufacture tape to enable its device technology. IBM’s typically be laid out on the direct access storage devices
first internally developed and manufactured tape, Series (DASD) that IBM had invented in the mid-1950s. Two
500, was introduced in 1967. This tape was produced at 3850 cartridges could hold 100 MB, which was the entire
the IBM Boulder facility until the introduction of the capacity of a 3330 disk pack. The 3850 cartridges could
Multi-System Tape (MST) in 1974. This gamma iron- hold 470 GB when fully configured, with almost 1000
oxide-based polyurethane and cross-linked vinyl resin 50-MB-capacity cylindrical cartridges [4]. This made it
binder system on a 1.5-mil-thick PET substrate became possible for a host system to use what appeared to be
the standard for durability and performance in the 1970s a huge disk storage device for a fraction of the cost.
and was arguably the best reel-to-reel computer tape ever The IBM 3850 was the first subsystem to provide a
made [3]. virtualized (or soft) image of a storage format that was 377

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER
 different from the actual physical storage target (as
presented to an application or storage management
software). Thus, the host and application software treated
the data as if it was stored on the DASD, rather than on
tape cartridges. The 3850 was also one of the earliest
devices to employ automation and fault management
within the device (independent of the host) [5–7].

The IBM 3480

The move to Tucson, Arizona, was announced in 1977,
and portions of the tape mission began the move shortly
thereafter. The move to this new site, specifically built to
develop and manufacture tape, tape heads, tape drives,
cards, and operational software and microcode, saw a
surge in new hires and transfers into tape from all over
IBM to join the team in Tucson.
The new site was officially dedicated in February 1980,
with more than 3500 people on the site, a number that
had swelled to more than 5000 employees by 1987. The
purpose of this effort was to introduce a series of
breakthroughs in tape storage technology. When the 3480
Figure 6 was announced on March 22, 1984, Al Rizzi, General
Products Division Tucson Laboratory Tape Products
The IBM 3480, introduced in 1984, standing in front of the IBM
manager, said, “Everything is new about this product. The
3420, which was the standard for reel-to-reel tapes prior to the 3480.
result is a tape subsystem that sets the industry standard.”
Tape storage was the primary mission of the IBM Tucson
Laboratory, and the 3480 announcement was a major
milestone for the site [1].
Two of the 3480’s radical new features were immediately
obvious: It used tape cartridges instead of reels, and it was
significantly smaller in size than the industry-standard IBM
3420 Model 11 10.5-inch-reel tape drive. While the 3420
placed 180 MB of data on a reel of tape and had an
uncompressed data rate of 1.25 MB/s, the 3480 had a capacity
of 200 MB in a 5.5-inch-square cartridge and a data rate
of 3 MB/s. The 3480 required less than half the floor space of
an equivalent installation of 3420s (Figures 6 and 7).
The 3480 introduced a number of technology
breakthroughs: the first thin-film and magnetorestrictive
(MR) head in a tape drive (and several years ahead of the
introduction of thin-film heads into disk drives) [8]; the
first 18-track head/channel in a tape drive; a new media
format (the half-inch data cartridge) on chromium dioxide
tape [9]; significant function and complex system error
logging enabled by major advances in embedded software
(microcode); and reduced power and space requirements
with significantly improved performance and reliability.
Vacuum columns, the essential component of the reel-to-
reel tape drives of the previous 25 years, were replaced in
the 3480 with direct reel-to-reel control. A significantly
Figure 7 improved and highly integrated write- and read-
The IBM 3480 200-MB cartridge compared with the 3420 Multi- equalization channel was introduced, and significant
System Tape 180-MB 10.5-inch reel. improvements in data error detection and correction
378 using adaptive cross parity (AXP) were achieved [10].

R. BRADSHAW AND C. SCHROEDER IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003
 To support the revolutionary new tape format, IBM Fibre Channel support with its introduction of the 3495
once again developed and manufactured a tape to support Tape Library Dataserver. It was a collaborative effort that
its new technology. With a customized binder formulation adapted existing robotics to a tape automation solution.
that exhibited superior durability and performance, the This large automated tape library was IBM’s first effort to
3480 cartridge system tape, or 3480 CST, became, within provide an automated tape solution since its 3850 MSS.
only a few years, the world standard for computer The 3495 was not an optimal solution, so the team
information storage on tape and forever changed the in Tucson started from scratch and, in a very short
expectations for tape reliability [11]. development time, delivered the more-efficient and
The 3480 product family was enhanced in 1989 with higher-capacity 3494 Tape Library Dataserver at the end
the introduction of the 36-track, 3490E drive and a new of 1993.
extended-length chromium dioxide tape. The 3490E DASD, or hard disk drives, also invented by IBM, and
provided 800 MB of storage in the same cartridge new storage devices using optical storage were being
format as the earlier 3480. The introduction of improved developed. New tape formats, disk arrays, optical devices,
recording capability (IDRC) further increased the capacity and technologies entered the marketplace and seemed
of the 3490E, to more than 2.4 GB, the highest data poised to relegate the use of tape storage to little more
capacity available at that time. than backup or archival repositories. Tape was beginning
The world was moving explosively into the computer to seem old and destined to be displaced by other
age. With the introduction of the personal computer, technologies. Hardly a year went by that did not see at
computers more powerful than the IBM 701 of 1952 least one pundit pronounce the impending demise of tape
were soon available to everyone. With the increased in the hierarchy of data storage. These assertions always
computerization of day-to-day transactions, an explosion assumed that innovations in tape storage had ceased to
of data was being generated—and stored. The sheer sustain its performance and cost advantages.
number of high-capacity tape cartridges in many
installations was becoming staggering; there were almost
IBM’s New Tape Program
100 million 3480/3490-compatible cartridges throughout
Dr. Paul Low, president of the IBM General Products
the world by the mid-1990s. It was becoming increasingly
Division, and Dr. Juri Matisoo, director of the IBM
imperative not only to improve tape capacity and
Almaden Research Center, convened a group early in
performance, but to provide automated tape storage
1988 to examine the viability of tape as a storage medium
and management.
for the future. One member of that task force, Dr. James
As noted earlier, with the announcement of the 3850
Eaton, recalled, “Early in the study it became apparent we
MSS in 1974, IBM offered the first automated tape
had to make radical improvements in linear tape-system
storage device. Support for the 3850 MSS was transferred
density and performance. Dramatic improvements were
to the Tucson Tape Laboratory from Boulder in the early
clearly possible. Tape technologies were available with
1980s, while manufacturing remained in Boulder during
over 100 times the density of the very successful IBM
the final years of 3850 availability. The revolutionary way 2
3480/3490 family of tape devices.”
of looking at tape in the hierarchy of storage initiated by
Although such existing tape technologies improved
the development of this pioneering tape automation
capacity, they did not offer the data rate required by the
device left a nucleus of talented and experienced people
applications on which IBM storage products were focused.
who created a number of innovations in data storage
management and architecture. The first cache-control The goal was to achieve both increased capacity and
device, the IBM 3880 Model 11/13, was developed in the increased data rate without sacrificing the proven
Tucson Laboratory in 1984, at least in part on the basis of reliability of the 3480/3490 products, which had become
experience gained with the 3850. This initial cache-control a world standard for information storage on tape. The
system led to successive advances in data storage control prevailing attitude among the group members at the
function and improved utilization of both IBM DASD and conclusion of the task force was that of excitement, with a
tape storage. This effort was very likely the beginning of reinvigorated focus on advancing IBM’s presence in tape
what is now called autonomic computing and virtualization. technology. Dr. Eaton summarized, “At the end of the
task force study, it was obvious that data storage tape
The IBM Tape Library Dataserver areal density could be increased by hundreds of times.
IBM had originally introduced automation with the We had to get busy to make it happen or risk losing the
3850 MSS. Then, in 1993, the IBM Tape Development business to whoever did!” Thus began IBM’s New Tape
2
Laboratory reentered the automation arena by combining Program (NTP) to explore novel directions for tape.
the strengths of world-class tape drives and control systems 2 Private communication with Dr. James Eaton, June 2002, with reference to
and advanced connectivity with ESCON* and then internal documents summarizing the Tape Task Force, May 1989. 379

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER
 The NTP effort explored a small-form-factor fast-access- single image to the host computer and operational
time tape format. What emerged from this advanced software. This seemingly simple function dramatically
technology group was a totally new, revolutionary device improved and optimized tape storage and the migration
using a timing-based servo, an advanced recording of large data sets [15].
channel, highly integrated signal processing, and a The 3590 was expanded in performance in 1998, with
completely new design for a low-mass head mounted on a the 3590E model doubling native capacity and improving
precision actuator assembly. The technology promised to performance. This was followed by the Magstar Extended
produce a small-form-factor device with a 5-GB native Length Cartridge media in 1999, which provided 40 GB of
capacity, a high data rate, and a very unique attribute for native capacity while preserving the existing automation
tape—rapid access to data (less than 11 seconds). This was and installed devices. The Enterprise tape family was thus
enabled by the use of a novel dual-reel midpoint load given four times more capacity and better than double the
cartridge. The technology implementation was delayed for data rate with the new device and tape combinations,
several years as the product focus shifted to address a while reliability and cost per gigabyte were improved as
replacement for the aging 3480/3490 Enterprise tape well [14].
products. Many of the key technology patents, however, Finally, the revolutionary tape architecture begun in the
were issued two or three years before the NTP product early 1990s as NTP was unveiled in 1996. The IBM 3570
was finally introduced [12]. Magstar MP with its 5-GB cartridge was the first small-
While a focused effort was underway to deliver the NTP form-factor midpoint-load tape device. It marked the
products, a very small team within the Tucson Laboratory introduction of a track-following timing-based servo and a
carried out a skunk-works project to move some of the number of innovations that leveraged the technology of
NTP technology into the aging 3480/3490 Enterprise IBM disk storage. In 1997, the capacity and performance
tape market. In 1995, a new family of devices, the IBM of the Magstar MP 3570 was extended to 7.5 GB with the
Magstar* 3590 tape drives, were introduced as the enhanced-capacity cartridge [16].
final products from the project, which had started so
3
slowly in the early 1990s [13]. This new family of high- An open-system solution
end IBM tape drives reset expectations for Enterprise An informal, internal push within the IBM Tape
tape and quickly became the platform upon which Development Laboratory began in late 1997 to move the
growing support for operating systems and applications exciting technology developed for the Magstar MP 3570
was based. into the existing product architectures. Eventually, this
The Magstar 3590 provided 10 GB of native capacity push evolved into an aggressive effort to establish, with
with a 9-MB/s data rate on a new servo-written metal- other storage industry leaders, a truly open-systems tape
particle tape in a 3480/3490-type cartridge, so it remained solution—Linear Tape-Open, or LTO. In less than two
compatible with existing automation devices. This marked years, this technology went from the laboratory to the
the first time that a new format was introduced while marketplace. The IBM Ultrium LTO drive brought with
preserving the investment in automation and addressing it new automation solutions and reemphasized the cost
the migration of tape technology to a new generation. As advantage of tape storage. It also provided a new impetus
a result, a tenfold reduction in the number of cartridges for the future. As the 20th century drew to a close, the
needed to store the existing 3480/3490 formats was IBM LTO was announced, and a new technology roadmap
achieved without requiring changes to the automation began for tape [17].
or significant disruption of customers’ operations. The innovations in data storage and processing on tape
The 3590 was a remarkable success and, within a few that had begun in the 1950s moved relatively slowly in its
years, became the new standard for Enterprise tape first 20 years, but the advances, particularly those related
storage [14]. to density and performance, were being achieved at an
A significant advance in performance and connectivity accelerating rate. The unfolding of a new century found
soon followed with the 3590 Magstar ESCON Control 100 GB of native capacity already available on a single
Unit Model A01 in 1996. The next year, the 3494 Model LTO cartridge. Magstar 3590 had just doubled the
B16 Virtual Tape Server (VTS) was introduced. This capacity of its full-function enterprise-level tape solution
device was the first to present a tape image to the host by introducing extended-length media. The performance
system (the 3850 MSS presented a disk image to the host) and reliability of both the initial 10-GB media and the
and greatly improved utilization of tape resources. The new extended-length media were attained through an
VTS optimized the filling of tapes while presenting a intensive co-development with leaders in the magnetic
3
tape industry. By 2002, the Magstar 3590 Model H had
Internal memo from John Gniewek to Steve Vogel, October 25, 1991 (IBM
380 Confidential). been introduced. It provided up to 60 GB of native

R. BRADSHAW AND C. SCHROEDER IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003
capacity and greatly improved reliability and stability to a Summary and conclusion
now mature enterprise tape format. IBM continues to push the boundaries of tape storage
Many of the breakthroughs in IBM disk storage in the technology. In 2002, the 50th anniversary of IBM’s
early 2000s made their way into tape and tape controllers, introduction of tape storage, one terabyte of uncompressed
including the ability to partition physical libraries into a information was successfully written and, more significantly,
number of logical libraries, Fibre Channel support, peer- read back successfully in a single half-inch tape cartridge
to-peer duplication of data, support of Linux** and other equivalent in size to the current LTO tape cartridge. The
Open Systems, and new connectivity options across a compression of one terabyte of data into a four-inch-wide by
growing number of storage platforms. At the same time, five-inch-long by one-inch-thick cartridge exceeds a density
the reliability, serviceability, and overall cost of IBM tape of one gigabit per square inch on the recorded media.
subsystems improved while tape technology in general was This advantage was achieved using advanced particulate
called upon to hold increasingly greater amounts of the tape technology coupled with improved high-density track
world’s data [18]. placement, made possible by utilizing a novel track-
The IBM LTO Model 3580 head was the first IBM following timing-based servo invented by IBM and the
tape drive recording head to break away from the older most advanced recording head technology used in linear
modular closed and contoured head designs that had tape drives. By leveraging the technology breakthroughs
carried the burden of linear tape drive recording for so pioneered by IBM for hard disk storage into tape storage
long. Given the proven capabilities of these disk head applications, one terabyte is only a milestone, not a
technologies, significant increases in tape track density barrier. The next 50 years look as good as the last 50
were possible. This improved head process allowed the years have been. We should expect to see capacity
mass of the head to be reduced while achieving synergy improving, substantial gains being made in data rate and
with disk head technology and process advances. reliability, and greater efficiencies in tape storage
A timing-based servo, invented by IBM, provides management.
very precise position information to the drive so that
interleaved bands of data can be written in a serpentine Acknowledgments
manner, eight tracks at a time, to construct 384 tracks The authors acknowledge the support of both active
capable of storing 100 GB of uncompressed data in the and retired IBM workers who provided reviews
first-generation LTO cartridge. Coupled with improved and commentary as well as copies of hard-to-find
media, actuator, and format innovations, significant documentation of the early pioneering efforts in tape
increases in the reliable placement of very narrow tracks storage. A special thanks goes to Wayne Winger, Max
on a thin tape were demonstrated with the first LTO tape Femmer, Bill Phillips, Billy Joe Mooney, John Carter, and
product. the IBM Tucson Communications Group, who found and
The IBM LTO family of tape drives will deliver increased digitized the photographs of the early hardware. No
storage capacity and performance for several generations. history of IBM tape could be made, furthermore,
Consistent with the LTO roadmap, a 200-GB-capacity without a sincere acknowledgment of the many
cartridge was announced in 2002. It delivers a significant people who hold the patents, designed the devices and
improvement in data rate from 15 MB/s to 35 MB/s. subsystems, integrated the software and functions, built
LTO storage currently provides more than several the products, sold IBM tape to customers, and serviced
hundred terabytes of uncompressed capacity in a and supported tape products all over the world for more
than 50 years, and who will continue to do so.
fully automated IBM tape library that takes up a space
equivalent to only eight five-drawer file cabinets. These
*Trademark or registered trademark of International Business
automated tape solutions are able to support multiple tape Machines Corporation.
formats in the same physical library separated from the
**Trademark or registered trademark of 3M Company,
host by distances of 100 or more kilometers. Additional General Electric Company, General Motors, and Linus
capacity increases, serviceability features and functions, Torvalds.
support of multiple tape formats within a single physical †
Ultrium, Linear Tape-Open, and LTO are registered
library, and logical partitioning of libraries are being trademarks of International Business Machines Corporation,
introduced to sustain the viability of tape storage well into Hewlett-Packard Corporation, and Seagate Corporation in the
United States, other countries, or both.
the future. IBM’s long history and experience with tape
and, indeed, all forms of digital information storage,
References
provides it with a unique ability to offer the seamless
1. C. Schroeder, “A History of Tape,” IBM Marketing
integration and management of information storage and Document, 1979; updated and reprinted in a second
retrieval that emerging on-demand businesses will require. perspective, “Celebrating Forty Years of Tape,” 1992. 381

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER
 2. E. W. Pugh, Building IBM: Shaping an Industry and Its 18. C. Brooks, A. Abete, M. Ceresoli, and J. Weingand,
Technology, The MIT Press, Cambridge, MA, 1995, “Using IBM LTO Ultrium with Open Systems,” IBM
pp. 167–182. Redbook SG24-6502-00 (January 2002).
3. J. P. Harris, W. B. Phillips, J. F. Wells, and W. D.
Winger, “Innovations in the Design of Magnetic Tape
Subsystems,” IBM J. Res. & Dev. 25, No. 5, 691– 699
Received February 26, 2003; accepted for publication
(1981).
March 13, 2003
4. J. P. Harris, R. S. Rohde, and N. K. Arter, “The IBM
3850 Mass Storage System: Design Aspects,” Proc. IEEE
63, No. 8, 1171–1176 (1978).
5. M. Bohl, Introduction to IBM Direct Access Storage
Devices, Science Research Association, Inc., Chicago,
1981, pp. 91–115.
6. S. S. Lavenberg and D. R. Slutz, “Regenerative
Simulation of a Queuing Model of an Automated Tape
Library,” IBM J. Res. & Dev. 19, No. 5, 463– 475 (1975).
7. W. G. Tuel, Jr., “An Analysis of Buffer Paging in Virtual
Storage Systems,” IBM J. Res. & Dev. 20, No. 5, 518 –520
(1976).
8. D. M. Cannon, W. R. Dempwolf, J. M. Schmalhorst, F. B.
Shelledy, and R. D. Silkensen, “Design and Performance
of a Magnetic Head for a High-Density Tape Drive,” IBM
J. Res. & Dev. 30, No. 3, 270 –277 (1986).
9. D. J. Winarski, W. W.-C. Chow, J. G. Bullock, F. B.
Froehlich, and T. G. Osterday, “Mechanical Design of the
Cartridge and Transport for the IBM 3480 Magnetic Tape
Subsystem,” IBM J. Res. & Dev. 30, No. 6, 635– 644
(1986).
10. A. M. Patel, “Adaptive Cross-Parity (AXP) Code for a
High-Density Magnetic Tape Subsystem,” IBM J. Res. &
Dev. 29, No. 6, 546 –562 (1985).
11. R. Bradshaw, B. Bhushan, C. Kalthoff, and M. Warne,
“Chemical and Mechanical Performance of Flexible
Magnetic Tape Containing Chromium Dioxide,” IBM J.
Res. & Dev. 30, No. 2, 203–216 (1986); R. L. Bradshaw
and T. W. Reid, “Archival Stability of IBM 3480/3490
Cartridge Tapes,” IEEE Trans. Magn. 27, No. 5, 4388 –
4395 (1991).
12. W. T. Comeaux, D. G. East, S. L. Felde, and D. W.
Johnson, “Tape Path Having Implicit Squeeze Bearing,”
U.S. Patent 5,284,308, February 1994; D. G. East, S. L.
Felde, P. Y. Hu, and G. S. Robles, “Tape Cartridge with
Mount Plate,” U.S. Patent 5,316,235, May 1994; S. L.
Felde and K. R. Shelley, “Machine and Human Readable
Label for Data Cartridge,” U.S. Patent 5,484,055, January
1996.
13. J. J. Gniewek and S. M. Vogel, “Influence of Technology
on Magnetic Tape Storage Device Characteristics,” NASA
Conf. Publ. 3295, Proceedings of the 4th NASA Goddard/
IEEE Joint Conference on Mass Storage Systems and
Technologies, 1995, pp. 237–351.
14. M. Massaki, “Magstar and IBM 3590 High Performance
Tape Subsystem: Multiplatform Implementation,” IBM
Redbook SG24-2594-02 (December 1996); M. Werner,
G. Gruber, J. Mitchell, C. Bauske, M. Oberholzer, and
R. Richards, “IBM Magstar Tape Products Family: A
Practical Guide,” IBM Redbook SG24-4632-03 (March
2000).
15. B. Kadleck, A. Bentley, W. Kessel, and A. Vulic,
“IBM TotalStorage Virtual Tape Server Planning,
Implementation and Monitoring,” IBM Redbook
SG24-2229-04 (September 2002).
16. M. Werner, J. Beveridge, G. Cormack, C. Coutts,
D. Lebreton, B. Manger, K. Nystrom, J.-J. Wattel, and
M. Werner, “Magstar MP 3575 Tape Library Dataserver:
Multiplatform Implementation,” IBM Redbook
SG24-4983-00 (April 1988).
17. M. Werner, C. Coutts, M. Marion, and N. Saavedra, “The
IBM LTO Ultrium Tape Libraries Guide,” IBM Redbook
382 SG24-5946-00 (November 2000).

R. BRADSHAW AND C. SCHROEDER IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003
Richard Bradshaw IBM Systems Group, 9000 South Rita
Road, Tucson, Arizona 85744 (rlb@us.ibm.com). Dr. Bradshaw
joined the Tape Development Group in the IBM General
Products Division after completing his Ph.D. degree in
synthetic organic/polymer chemistry in 1978. He spent one
year in the IBM Boulder Laboratory before moving to the
new IBM Development Laboratory in Tucson, Arizona. Since
then he has been involved in magnetic media formulation,
characterization, and development, having worked as a major
contributor on the tapes used in the IBM 3480, the 3490E,
the 3570 MP, 3590, LTO, and a number of other storage
products. Dr. Bradshaw is actively involved in ongoing
research activities both within and outside IBM that are
intended to advance the technology of tape storage well into
the future.

Carl Schroeder IBM Systems Group, 5600 Cottle Road,

San Jose, California 95193 (carljs@us.ibm.com). Mr. Schroeder
joined IBM in 1978. He has worked in speech writing, press
relations, internal communications, marketing communications,
field communications, and as a systems engineer on the Chrysler
team; he was responsible for creating the public IBM Storage
Web site, www.ibm.com/storage. He researched and wrote
the first version of the History of Tape for the dedication
of the Tucson site in 1980 and updated it for the 40th
anniversary of tape in 1992.

383

IBM J. RES. & DEV. VOL. 47 NO. 4 JULY 2003 R. BRADSHAW AND C. SCHROEDER